Back to all discussions

  •   about 2 years ago

clarification on testing for judges

Hello,
In the project overview it says "Include at least one test script for your Project". Could you please clarify what exactly is meant by this?
Are we supposed to implement the conversational AI agent into an working app (like a website or a Messenger like Telegram)?
Or should we provide a link to a repository?

Thank you for clarifying.

  • 1 comment

  • Manager   •   about 2 years ago

    Hi there,

    Apologies for the delay. The Google Team is very busy but alas we have an answer :)

    We’re looking for a document such as a PDF that walks a user or tester through recommended prompts so they can experience the agent themselves. Two options would suffice:
    [Preferred] serve agent on an endpoint, indicated endpoint in the submission, and provide recommended conversation starters or prompts that a new user or tester could use to demo the abilities and features of your AI Agent.
    [Also Ok] If unable to serve the agent to other users, submit the same as above but also include screenshots of the responses your AI Agent outputs in the UI using the Preview Agent right-hand sidebar.

Comments are closed.